• Home
  • History
  • Annotate
  • Line#
  • Navigate
  • Raw
  • Download
  • only in /asuswrt-rt-n18u-9.0.0.4.380.2695/release/src-rt-6.x.4708/linux/linux-2.6.36/fs/ext4/
1/*
2 * linux/fs/ext4/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 *  suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
11 *  Red Hat Inc.
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 *  and Andreas Gruenbacher <agruen@suse.de>.
14 */
15
16/*
17 * Extended attributes are stored directly in inodes (on file systems with
18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
19 * field contains the block number if an inode uses an additional block. All
20 * attributes must fit in the inode and one additional block. Blocks that
21 * contain the identical set of attributes may be shared among several inodes.
22 * Identical blocks are detected by keeping a cache of blocks that have
23 * recently been accessed.
24 *
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
27 *
28 *   +------------------+
29 *   | header           |
30 *   | entry 1          | |
31 *   | entry 2          | | growing downwards
32 *   | entry 3          | v
33 *   | four null bytes  |
34 *   | . . .            |
35 *   | value 1          | ^
36 *   | value 3          | | growing upwards
37 *   | value 2          | |
38 *   +------------------+
39 *
40 * The header is followed by multiple entry descriptors. In disk blocks, the
41 * entry descriptors are kept sorted. In inodes, they are unsorted. The
42 * attribute values are aligned to the end of the block in no specific order.
43 *
44 * Locking strategy
45 * ----------------
46 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
50 * by the buffer lock.
51 */
52
53#include <linux/init.h>
54#include <linux/fs.h>
55#include <linux/slab.h>
56#include <linux/mbcache.h>
57#include <linux/quotaops.h>
58#include <linux/rwsem.h>
59#include "ext4_jbd2.h"
60#include "ext4.h"
61#include "xattr.h"
62#include "acl.h"
63
64#define BHDR(bh) ((struct ext4_xattr_header *)((bh)->b_data))
65#define ENTRY(ptr) ((struct ext4_xattr_entry *)(ptr))
66#define BFIRST(bh) ENTRY(BHDR(bh)+1)
67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
68
69#ifdef EXT4_XATTR_DEBUG
70# define ea_idebug(inode, f...) do { \
71		printk(KERN_DEBUG "inode %s:%lu: ", \
72			inode->i_sb->s_id, inode->i_ino); \
73		printk(f); \
74		printk("\n"); \
75	} while (0)
76# define ea_bdebug(bh, f...) do { \
77		char b[BDEVNAME_SIZE]; \
78		printk(KERN_DEBUG "block %s:%lu: ", \
79			bdevname(bh->b_bdev, b), \
80			(unsigned long) bh->b_blocknr); \
81		printk(f); \
82		printk("\n"); \
83	} while (0)
84#else
85# define ea_idebug(f...)
86# define ea_bdebug(f...)
87#endif
88
89static void ext4_xattr_cache_insert(struct buffer_head *);
90static struct buffer_head *ext4_xattr_cache_find(struct inode *,
91						 struct ext4_xattr_header *,
92						 struct mb_cache_entry **);
93static void ext4_xattr_rehash(struct ext4_xattr_header *,
94			      struct ext4_xattr_entry *);
95static int ext4_xattr_list(struct dentry *dentry, char *buffer,
96			   size_t buffer_size);
97
98static struct mb_cache *ext4_xattr_cache;
99
100static const struct xattr_handler *ext4_xattr_handler_map[] = {
101	[EXT4_XATTR_INDEX_USER]		     = &ext4_xattr_user_handler,
102#ifdef CONFIG_EXT4_FS_POSIX_ACL
103	[EXT4_XATTR_INDEX_POSIX_ACL_ACCESS]  = &ext4_xattr_acl_access_handler,
104	[EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
105#endif
106	[EXT4_XATTR_INDEX_TRUSTED]	     = &ext4_xattr_trusted_handler,
107#ifdef CONFIG_EXT4_FS_SECURITY
108	[EXT4_XATTR_INDEX_SECURITY]	     = &ext4_xattr_security_handler,
109#endif
110};
111
112const struct xattr_handler *ext4_xattr_handlers[] = {
113	&ext4_xattr_user_handler,
114	&ext4_xattr_trusted_handler,
115#ifdef CONFIG_EXT4_FS_POSIX_ACL
116	&ext4_xattr_acl_access_handler,
117	&ext4_xattr_acl_default_handler,
118#endif
119#ifdef CONFIG_EXT4_FS_SECURITY
120	&ext4_xattr_security_handler,
121#endif
122	NULL
123};
124
125static inline const struct xattr_handler *
126ext4_xattr_handler(int name_index)
127{
128	const struct xattr_handler *handler = NULL;
129
130	if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
131		handler = ext4_xattr_handler_map[name_index];
132	return handler;
133}
134
135/*
136 * Inode operation listxattr()
137 *
138 * dentry->d_inode->i_mutex: don't care
139 */
140ssize_t
141ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
142{
143	return ext4_xattr_list(dentry, buffer, size);
144}
145
146static int
147ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
148{
149	while (!IS_LAST_ENTRY(entry)) {
150		struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
151		if ((void *)next >= end)
152			return -EIO;
153		entry = next;
154	}
155	return 0;
156}
157
158static inline int
159ext4_xattr_check_block(struct buffer_head *bh)
160{
161	int error;
162
163	if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
164	    BHDR(bh)->h_blocks != cpu_to_le32(1))
165		return -EIO;
166	error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
167	return error;
168}
169
170static inline int
171ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
172{
173	size_t value_size = le32_to_cpu(entry->e_value_size);
174
175	if (entry->e_value_block != 0 || value_size > size ||
176	    le16_to_cpu(entry->e_value_offs) + value_size > size)
177		return -EIO;
178	return 0;
179}
180
181static int
182ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
183		      const char *name, size_t size, int sorted)
184{
185	struct ext4_xattr_entry *entry;
186	size_t name_len;
187	int cmp = 1;
188
189	if (name == NULL)
190		return -EINVAL;
191	name_len = strlen(name);
192	entry = *pentry;
193	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
194		cmp = name_index - entry->e_name_index;
195		if (!cmp)
196			cmp = name_len - entry->e_name_len;
197		if (!cmp)
198			cmp = memcmp(name, entry->e_name, name_len);
199		if (cmp <= 0 && (sorted || cmp == 0))
200			break;
201	}
202	*pentry = entry;
203	if (!cmp && ext4_xattr_check_entry(entry, size))
204			return -EIO;
205	return cmp ? -ENODATA : 0;
206}
207
208static int
209ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
210		     void *buffer, size_t buffer_size)
211{
212	struct buffer_head *bh = NULL;
213	struct ext4_xattr_entry *entry;
214	size_t size;
215	int error;
216
217	ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
218		  name_index, name, buffer, (long)buffer_size);
219
220	error = -ENODATA;
221	if (!EXT4_I(inode)->i_file_acl)
222		goto cleanup;
223	ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
224	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
225	if (!bh)
226		goto cleanup;
227	ea_bdebug(bh, "b_count=%d, refcount=%d",
228		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
229	if (ext4_xattr_check_block(bh)) {
230bad_block:
231		EXT4_ERROR_INODE(inode, "bad block %llu",
232				 EXT4_I(inode)->i_file_acl);
233		error = -EIO;
234		goto cleanup;
235	}
236	ext4_xattr_cache_insert(bh);
237	entry = BFIRST(bh);
238	error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
239	if (error == -EIO)
240		goto bad_block;
241	if (error)
242		goto cleanup;
243	size = le32_to_cpu(entry->e_value_size);
244	if (buffer) {
245		error = -ERANGE;
246		if (size > buffer_size)
247			goto cleanup;
248		memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
249		       size);
250	}
251	error = size;
252
253cleanup:
254	brelse(bh);
255	return error;
256}
257
258static int
259ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
260		     void *buffer, size_t buffer_size)
261{
262	struct ext4_xattr_ibody_header *header;
263	struct ext4_xattr_entry *entry;
264	struct ext4_inode *raw_inode;
265	struct ext4_iloc iloc;
266	size_t size;
267	void *end;
268	int error;
269
270	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
271		return -ENODATA;
272	error = ext4_get_inode_loc(inode, &iloc);
273	if (error)
274		return error;
275	raw_inode = ext4_raw_inode(&iloc);
276	header = IHDR(inode, raw_inode);
277	entry = IFIRST(header);
278	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
279	error = ext4_xattr_check_names(entry, end);
280	if (error)
281		goto cleanup;
282	error = ext4_xattr_find_entry(&entry, name_index, name,
283				      end - (void *)entry, 0);
284	if (error)
285		goto cleanup;
286	size = le32_to_cpu(entry->e_value_size);
287	if (buffer) {
288		error = -ERANGE;
289		if (size > buffer_size)
290			goto cleanup;
291		memcpy(buffer, (void *)IFIRST(header) +
292		       le16_to_cpu(entry->e_value_offs), size);
293	}
294	error = size;
295
296cleanup:
297	brelse(iloc.bh);
298	return error;
299}
300
301/*
302 * ext4_xattr_get()
303 *
304 * Copy an extended attribute into the buffer
305 * provided, or compute the buffer size required.
306 * Buffer is NULL to compute the size of the buffer required.
307 *
308 * Returns a negative error number on failure, or the number of bytes
309 * used / required on success.
310 */
311int
312ext4_xattr_get(struct inode *inode, int name_index, const char *name,
313	       void *buffer, size_t buffer_size)
314{
315	int error;
316
317	down_read(&EXT4_I(inode)->xattr_sem);
318	error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
319				     buffer_size);
320	if (error == -ENODATA)
321		error = ext4_xattr_block_get(inode, name_index, name, buffer,
322					     buffer_size);
323	up_read(&EXT4_I(inode)->xattr_sem);
324	return error;
325}
326
327static int
328ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
329			char *buffer, size_t buffer_size)
330{
331	size_t rest = buffer_size;
332
333	for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
334		const struct xattr_handler *handler =
335			ext4_xattr_handler(entry->e_name_index);
336
337		if (handler) {
338			size_t size = handler->list(dentry, buffer, rest,
339						    entry->e_name,
340						    entry->e_name_len,
341						    handler->flags);
342			if (buffer) {
343				if (size > rest)
344					return -ERANGE;
345				buffer += size;
346			}
347			rest -= size;
348		}
349	}
350	return buffer_size - rest;
351}
352
353static int
354ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
355{
356	struct inode *inode = dentry->d_inode;
357	struct buffer_head *bh = NULL;
358	int error;
359
360	ea_idebug(inode, "buffer=%p, buffer_size=%ld",
361		  buffer, (long)buffer_size);
362
363	error = 0;
364	if (!EXT4_I(inode)->i_file_acl)
365		goto cleanup;
366	ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
367	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
368	error = -EIO;
369	if (!bh)
370		goto cleanup;
371	ea_bdebug(bh, "b_count=%d, refcount=%d",
372		atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
373	if (ext4_xattr_check_block(bh)) {
374		EXT4_ERROR_INODE(inode, "bad block %llu",
375				 EXT4_I(inode)->i_file_acl);
376		error = -EIO;
377		goto cleanup;
378	}
379	ext4_xattr_cache_insert(bh);
380	error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
381
382cleanup:
383	brelse(bh);
384
385	return error;
386}
387
388static int
389ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
390{
391	struct inode *inode = dentry->d_inode;
392	struct ext4_xattr_ibody_header *header;
393	struct ext4_inode *raw_inode;
394	struct ext4_iloc iloc;
395	void *end;
396	int error;
397
398	if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
399		return 0;
400	error = ext4_get_inode_loc(inode, &iloc);
401	if (error)
402		return error;
403	raw_inode = ext4_raw_inode(&iloc);
404	header = IHDR(inode, raw_inode);
405	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
406	error = ext4_xattr_check_names(IFIRST(header), end);
407	if (error)
408		goto cleanup;
409	error = ext4_xattr_list_entries(dentry, IFIRST(header),
410					buffer, buffer_size);
411
412cleanup:
413	brelse(iloc.bh);
414	return error;
415}
416
417/*
418 * ext4_xattr_list()
419 *
420 * Copy a list of attribute names into the buffer
421 * provided, or compute the buffer size required.
422 * Buffer is NULL to compute the size of the buffer required.
423 *
424 * Returns a negative error number on failure, or the number of bytes
425 * used / required on success.
426 */
427static int
428ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
429{
430	int i_error, b_error;
431
432	down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
433	i_error = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
434	if (i_error < 0) {
435		b_error = 0;
436	} else {
437		if (buffer) {
438			buffer += i_error;
439			buffer_size -= i_error;
440		}
441		b_error = ext4_xattr_block_list(dentry, buffer, buffer_size);
442		if (b_error < 0)
443			i_error = 0;
444	}
445	up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
446	return i_error + b_error;
447}
448
449/*
450 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
451 * not set, set it.
452 */
453static void ext4_xattr_update_super_block(handle_t *handle,
454					  struct super_block *sb)
455{
456	if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
457		return;
458
459	if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
460		EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
461		ext4_handle_dirty_super(handle, sb);
462	}
463}
464
465/*
466 * Release the xattr block BH: If the reference count is > 1, decrement
467 * it; otherwise free the block.
468 */
469static void
470ext4_xattr_release_block(handle_t *handle, struct inode *inode,
471			 struct buffer_head *bh)
472{
473	struct mb_cache_entry *ce = NULL;
474	int error = 0;
475
476	ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
477	error = ext4_journal_get_write_access(handle, bh);
478	if (error)
479		goto out;
480
481	lock_buffer(bh);
482	if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
483		ea_bdebug(bh, "refcount now=0; freeing");
484		if (ce)
485			mb_cache_entry_free(ce);
486		get_bh(bh);
487		ext4_free_blocks(handle, inode, bh, 0, 1,
488				 EXT4_FREE_BLOCKS_METADATA |
489				 EXT4_FREE_BLOCKS_FORGET);
490	} else {
491		le32_add_cpu(&BHDR(bh)->h_refcount, -1);
492		error = ext4_handle_dirty_metadata(handle, inode, bh);
493		if (IS_SYNC(inode))
494			ext4_handle_sync(handle);
495		dquot_free_block(inode, 1);
496		ea_bdebug(bh, "refcount now=%d; releasing",
497			  le32_to_cpu(BHDR(bh)->h_refcount));
498		if (ce)
499			mb_cache_entry_release(ce);
500	}
501	unlock_buffer(bh);
502out:
503	ext4_std_error(inode->i_sb, error);
504	return;
505}
506
507/*
508 * Find the available free space for EAs. This also returns the total number of
509 * bytes used by EA entries.
510 */
511static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
512				    size_t *min_offs, void *base, int *total)
513{
514	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
515		*total += EXT4_XATTR_LEN(last->e_name_len);
516		if (!last->e_value_block && last->e_value_size) {
517			size_t offs = le16_to_cpu(last->e_value_offs);
518			if (offs < *min_offs)
519				*min_offs = offs;
520		}
521	}
522	return (*min_offs - ((void *)last - base) - sizeof(__u32));
523}
524
525struct ext4_xattr_info {
526	int name_index;
527	const char *name;
528	const void *value;
529	size_t value_len;
530};
531
532struct ext4_xattr_search {
533	struct ext4_xattr_entry *first;
534	void *base;
535	void *end;
536	struct ext4_xattr_entry *here;
537	int not_found;
538};
539
540static int
541ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
542{
543	struct ext4_xattr_entry *last;
544	size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
545
546	/* Compute min_offs and last. */
547	last = s->first;
548	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
549		if (!last->e_value_block && last->e_value_size) {
550			size_t offs = le16_to_cpu(last->e_value_offs);
551			if (offs < min_offs)
552				min_offs = offs;
553		}
554	}
555	free = min_offs - ((void *)last - s->base) - sizeof(__u32);
556	if (!s->not_found) {
557		if (!s->here->e_value_block && s->here->e_value_size) {
558			size_t size = le32_to_cpu(s->here->e_value_size);
559			free += EXT4_XATTR_SIZE(size);
560		}
561		free += EXT4_XATTR_LEN(name_len);
562	}
563	if (i->value) {
564		if (free < EXT4_XATTR_SIZE(i->value_len) ||
565		    free < EXT4_XATTR_LEN(name_len) +
566			   EXT4_XATTR_SIZE(i->value_len))
567			return -ENOSPC;
568	}
569
570	if (i->value && s->not_found) {
571		/* Insert the new name. */
572		size_t size = EXT4_XATTR_LEN(name_len);
573		size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
574		memmove((void *)s->here + size, s->here, rest);
575		memset(s->here, 0, size);
576		s->here->e_name_index = i->name_index;
577		s->here->e_name_len = name_len;
578		memcpy(s->here->e_name, i->name, name_len);
579	} else {
580		if (!s->here->e_value_block && s->here->e_value_size) {
581			void *first_val = s->base + min_offs;
582			size_t offs = le16_to_cpu(s->here->e_value_offs);
583			void *val = s->base + offs;
584			size_t size = EXT4_XATTR_SIZE(
585				le32_to_cpu(s->here->e_value_size));
586
587			if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
588				/* The old and the new value have the same
589				   size. Just replace. */
590				s->here->e_value_size =
591					cpu_to_le32(i->value_len);
592				memset(val + size - EXT4_XATTR_PAD, 0,
593				       EXT4_XATTR_PAD); /* Clear pad bytes. */
594				memcpy(val, i->value, i->value_len);
595				return 0;
596			}
597
598			/* Remove the old value. */
599			memmove(first_val + size, first_val, val - first_val);
600			memset(first_val, 0, size);
601			s->here->e_value_size = 0;
602			s->here->e_value_offs = 0;
603			min_offs += size;
604
605			/* Adjust all value offsets. */
606			last = s->first;
607			while (!IS_LAST_ENTRY(last)) {
608				size_t o = le16_to_cpu(last->e_value_offs);
609				if (!last->e_value_block &&
610				    last->e_value_size && o < offs)
611					last->e_value_offs =
612						cpu_to_le16(o + size);
613				last = EXT4_XATTR_NEXT(last);
614			}
615		}
616		if (!i->value) {
617			/* Remove the old name. */
618			size_t size = EXT4_XATTR_LEN(name_len);
619			last = ENTRY((void *)last - size);
620			memmove(s->here, (void *)s->here + size,
621				(void *)last - (void *)s->here + sizeof(__u32));
622			memset(last, 0, size);
623		}
624	}
625
626	if (i->value) {
627		/* Insert the new value. */
628		s->here->e_value_size = cpu_to_le32(i->value_len);
629		if (i->value_len) {
630			size_t size = EXT4_XATTR_SIZE(i->value_len);
631			void *val = s->base + min_offs - size;
632			s->here->e_value_offs = cpu_to_le16(min_offs - size);
633			memset(val + size - EXT4_XATTR_PAD, 0,
634			       EXT4_XATTR_PAD); /* Clear the pad bytes. */
635			memcpy(val, i->value, i->value_len);
636		}
637	}
638	return 0;
639}
640
641struct ext4_xattr_block_find {
642	struct ext4_xattr_search s;
643	struct buffer_head *bh;
644};
645
646static int
647ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
648		      struct ext4_xattr_block_find *bs)
649{
650	struct super_block *sb = inode->i_sb;
651	int error;
652
653	ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
654		  i->name_index, i->name, i->value, (long)i->value_len);
655
656	if (EXT4_I(inode)->i_file_acl) {
657		/* The inode already has an extended attribute block. */
658		bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
659		error = -EIO;
660		if (!bs->bh)
661			goto cleanup;
662		ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
663			atomic_read(&(bs->bh->b_count)),
664			le32_to_cpu(BHDR(bs->bh)->h_refcount));
665		if (ext4_xattr_check_block(bs->bh)) {
666			EXT4_ERROR_INODE(inode, "bad block %llu",
667					 EXT4_I(inode)->i_file_acl);
668			error = -EIO;
669			goto cleanup;
670		}
671		/* Find the named attribute. */
672		bs->s.base = BHDR(bs->bh);
673		bs->s.first = BFIRST(bs->bh);
674		bs->s.end = bs->bh->b_data + bs->bh->b_size;
675		bs->s.here = bs->s.first;
676		error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
677					      i->name, bs->bh->b_size, 1);
678		if (error && error != -ENODATA)
679			goto cleanup;
680		bs->s.not_found = error;
681	}
682	error = 0;
683
684cleanup:
685	return error;
686}
687
688static int
689ext4_xattr_block_set(handle_t *handle, struct inode *inode,
690		     struct ext4_xattr_info *i,
691		     struct ext4_xattr_block_find *bs)
692{
693	struct super_block *sb = inode->i_sb;
694	struct buffer_head *new_bh = NULL;
695	struct ext4_xattr_search *s = &bs->s;
696	struct mb_cache_entry *ce = NULL;
697	int error = 0;
698
699#define header(x) ((struct ext4_xattr_header *)(x))
700
701	if (i->value && i->value_len > sb->s_blocksize)
702		return -ENOSPC;
703	if (s->base) {
704		ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
705					bs->bh->b_blocknr);
706		error = ext4_journal_get_write_access(handle, bs->bh);
707		if (error)
708			goto cleanup;
709		lock_buffer(bs->bh);
710
711		if (header(s->base)->h_refcount == cpu_to_le32(1)) {
712			if (ce) {
713				mb_cache_entry_free(ce);
714				ce = NULL;
715			}
716			ea_bdebug(bs->bh, "modifying in-place");
717			error = ext4_xattr_set_entry(i, s);
718			if (!error) {
719				if (!IS_LAST_ENTRY(s->first))
720					ext4_xattr_rehash(header(s->base),
721							  s->here);
722				ext4_xattr_cache_insert(bs->bh);
723			}
724			unlock_buffer(bs->bh);
725			if (error == -EIO)
726				goto bad_block;
727			if (!error)
728				error = ext4_handle_dirty_metadata(handle,
729								   inode,
730								   bs->bh);
731			if (error)
732				goto cleanup;
733			goto inserted;
734		} else {
735			int offset = (char *)s->here - bs->bh->b_data;
736
737			unlock_buffer(bs->bh);
738			jbd2_journal_release_buffer(handle, bs->bh);
739			if (ce) {
740				mb_cache_entry_release(ce);
741				ce = NULL;
742			}
743			ea_bdebug(bs->bh, "cloning");
744			s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
745			error = -ENOMEM;
746			if (s->base == NULL)
747				goto cleanup;
748			memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
749			s->first = ENTRY(header(s->base)+1);
750			header(s->base)->h_refcount = cpu_to_le32(1);
751			s->here = ENTRY(s->base + offset);
752			s->end = s->base + bs->bh->b_size;
753		}
754	} else {
755		/* Allocate a buffer where we construct the new block. */
756		s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
757		/* assert(header == s->base) */
758		error = -ENOMEM;
759		if (s->base == NULL)
760			goto cleanup;
761		header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
762		header(s->base)->h_blocks = cpu_to_le32(1);
763		header(s->base)->h_refcount = cpu_to_le32(1);
764		s->first = ENTRY(header(s->base)+1);
765		s->here = ENTRY(header(s->base)+1);
766		s->end = s->base + sb->s_blocksize;
767	}
768
769	error = ext4_xattr_set_entry(i, s);
770	if (error == -EIO)
771		goto bad_block;
772	if (error)
773		goto cleanup;
774	if (!IS_LAST_ENTRY(s->first))
775		ext4_xattr_rehash(header(s->base), s->here);
776
777inserted:
778	if (!IS_LAST_ENTRY(s->first)) {
779		new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
780		if (new_bh) {
781			/* We found an identical block in the cache. */
782			if (new_bh == bs->bh)
783				ea_bdebug(new_bh, "keeping");
784			else {
785				/* The old block is released after updating
786				   the inode. */
787				error = dquot_alloc_block(inode, 1);
788				if (error)
789					goto cleanup;
790				error = ext4_journal_get_write_access(handle,
791								      new_bh);
792				if (error)
793					goto cleanup_dquot;
794				lock_buffer(new_bh);
795				le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
796				ea_bdebug(new_bh, "reusing; refcount now=%d",
797					le32_to_cpu(BHDR(new_bh)->h_refcount));
798				unlock_buffer(new_bh);
799				error = ext4_handle_dirty_metadata(handle,
800								   inode,
801								   new_bh);
802				if (error)
803					goto cleanup_dquot;
804			}
805			mb_cache_entry_release(ce);
806			ce = NULL;
807		} else if (bs->bh && s->base == bs->bh->b_data) {
808			/* We were modifying this block in-place. */
809			ea_bdebug(bs->bh, "keeping this block");
810			new_bh = bs->bh;
811			get_bh(new_bh);
812		} else {
813			/* We need to allocate a new block */
814			ext4_fsblk_t goal, block;
815
816			goal = ext4_group_first_block_no(sb,
817						EXT4_I(inode)->i_block_group);
818
819			/* non-extent files can't have physical blocks past 2^32 */
820			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
821				goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
822
823			block = ext4_new_meta_blocks(handle, inode,
824						  goal, NULL, &error);
825			if (error)
826				goto cleanup;
827
828			if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
829				BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
830
831			ea_idebug(inode, "creating block %d", block);
832
833			new_bh = sb_getblk(sb, block);
834			if (!new_bh) {
835getblk_failed:
836				ext4_free_blocks(handle, inode, 0, block, 1,
837						 EXT4_FREE_BLOCKS_METADATA);
838				error = -EIO;
839				goto cleanup;
840			}
841			lock_buffer(new_bh);
842			error = ext4_journal_get_create_access(handle, new_bh);
843			if (error) {
844				unlock_buffer(new_bh);
845				goto getblk_failed;
846			}
847			memcpy(new_bh->b_data, s->base, new_bh->b_size);
848			set_buffer_uptodate(new_bh);
849			unlock_buffer(new_bh);
850			ext4_xattr_cache_insert(new_bh);
851			error = ext4_handle_dirty_metadata(handle,
852							   inode, new_bh);
853			if (error)
854				goto cleanup;
855		}
856	}
857
858	/* Update the inode. */
859	EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
860
861	/* Drop the previous xattr block. */
862	if (bs->bh && bs->bh != new_bh)
863		ext4_xattr_release_block(handle, inode, bs->bh);
864	error = 0;
865
866cleanup:
867	if (ce)
868		mb_cache_entry_release(ce);
869	brelse(new_bh);
870	if (!(bs->bh && s->base == bs->bh->b_data))
871		kfree(s->base);
872
873	return error;
874
875cleanup_dquot:
876	dquot_free_block(inode, 1);
877	goto cleanup;
878
879bad_block:
880	EXT4_ERROR_INODE(inode, "bad block %llu",
881			 EXT4_I(inode)->i_file_acl);
882	goto cleanup;
883
884#undef header
885}
886
887struct ext4_xattr_ibody_find {
888	struct ext4_xattr_search s;
889	struct ext4_iloc iloc;
890};
891
892static int
893ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
894		      struct ext4_xattr_ibody_find *is)
895{
896	struct ext4_xattr_ibody_header *header;
897	struct ext4_inode *raw_inode;
898	int error;
899
900	if (EXT4_I(inode)->i_extra_isize == 0)
901		return 0;
902	raw_inode = ext4_raw_inode(&is->iloc);
903	header = IHDR(inode, raw_inode);
904	is->s.base = is->s.first = IFIRST(header);
905	is->s.here = is->s.first;
906	is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
907	if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
908		error = ext4_xattr_check_names(IFIRST(header), is->s.end);
909		if (error)
910			return error;
911		/* Find the named attribute. */
912		error = ext4_xattr_find_entry(&is->s.here, i->name_index,
913					      i->name, is->s.end -
914					      (void *)is->s.base, 0);
915		if (error && error != -ENODATA)
916			return error;
917		is->s.not_found = error;
918	}
919	return 0;
920}
921
922static int
923ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
924		     struct ext4_xattr_info *i,
925		     struct ext4_xattr_ibody_find *is)
926{
927	struct ext4_xattr_ibody_header *header;
928	struct ext4_xattr_search *s = &is->s;
929	int error;
930
931	if (EXT4_I(inode)->i_extra_isize == 0)
932		return -ENOSPC;
933	error = ext4_xattr_set_entry(i, s);
934	if (error)
935		return error;
936	header = IHDR(inode, ext4_raw_inode(&is->iloc));
937	if (!IS_LAST_ENTRY(s->first)) {
938		header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
939		ext4_set_inode_state(inode, EXT4_STATE_XATTR);
940	} else {
941		header->h_magic = cpu_to_le32(0);
942		ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
943	}
944	return 0;
945}
946
947/*
948 * ext4_xattr_set_handle()
949 *
950 * Create, replace or remove an extended attribute for this inode. Buffer
951 * is NULL to remove an existing extended attribute, and non-NULL to
952 * either replace an existing extended attribute, or create a new extended
953 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
954 * specify that an extended attribute must exist and must not exist
955 * previous to the call, respectively.
956 *
957 * Returns 0, or a negative error number on failure.
958 */
959int
960ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
961		      const char *name, const void *value, size_t value_len,
962		      int flags)
963{
964	struct ext4_xattr_info i = {
965		.name_index = name_index,
966		.name = name,
967		.value = value,
968		.value_len = value_len,
969
970	};
971	struct ext4_xattr_ibody_find is = {
972		.s = { .not_found = -ENODATA, },
973	};
974	struct ext4_xattr_block_find bs = {
975		.s = { .not_found = -ENODATA, },
976	};
977	unsigned long no_expand;
978	int error;
979
980	if (!name)
981		return -EINVAL;
982	if (strlen(name) > 255)
983		return -ERANGE;
984	down_write(&EXT4_I(inode)->xattr_sem);
985	no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
986	ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
987
988	error = ext4_get_inode_loc(inode, &is.iloc);
989	if (error)
990		goto cleanup;
991
992	error = ext4_journal_get_write_access(handle, is.iloc.bh);
993	if (error)
994		goto cleanup;
995
996	if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
997		struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
998		memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
999		ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1000	}
1001
1002	error = ext4_xattr_ibody_find(inode, &i, &is);
1003	if (error)
1004		goto cleanup;
1005	if (is.s.not_found)
1006		error = ext4_xattr_block_find(inode, &i, &bs);
1007	if (error)
1008		goto cleanup;
1009	if (is.s.not_found && bs.s.not_found) {
1010		error = -ENODATA;
1011		if (flags & XATTR_REPLACE)
1012			goto cleanup;
1013		error = 0;
1014		if (!value)
1015			goto cleanup;
1016	} else {
1017		error = -EEXIST;
1018		if (flags & XATTR_CREATE)
1019			goto cleanup;
1020	}
1021	if (!value) {
1022		if (!is.s.not_found)
1023			error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1024		else if (!bs.s.not_found)
1025			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1026	} else {
1027		error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1028		if (!error && !bs.s.not_found) {
1029			i.value = NULL;
1030			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1031		} else if (error == -ENOSPC) {
1032			if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1033				error = ext4_xattr_block_find(inode, &i, &bs);
1034				if (error)
1035					goto cleanup;
1036			}
1037			error = ext4_xattr_block_set(handle, inode, &i, &bs);
1038			if (error)
1039				goto cleanup;
1040			if (!is.s.not_found) {
1041				i.value = NULL;
1042				error = ext4_xattr_ibody_set(handle, inode, &i,
1043							     &is);
1044			}
1045		}
1046	}
1047	if (!error) {
1048		ext4_xattr_update_super_block(handle, inode->i_sb);
1049		inode->i_ctime = ext4_current_time(inode);
1050		if (!value)
1051			ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1052		error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1053		/*
1054		 * The bh is consumed by ext4_mark_iloc_dirty, even with
1055		 * error != 0.
1056		 */
1057		is.iloc.bh = NULL;
1058		if (IS_SYNC(inode))
1059			ext4_handle_sync(handle);
1060	}
1061
1062cleanup:
1063	brelse(is.iloc.bh);
1064	brelse(bs.bh);
1065	if (no_expand == 0)
1066		ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1067	up_write(&EXT4_I(inode)->xattr_sem);
1068	return error;
1069}
1070
1071/*
1072 * ext4_xattr_set()
1073 *
1074 * Like ext4_xattr_set_handle, but start from an inode. This extended
1075 * attribute modification is a filesystem transaction by itself.
1076 *
1077 * Returns 0, or a negative error number on failure.
1078 */
1079int
1080ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1081	       const void *value, size_t value_len, int flags)
1082{
1083	handle_t *handle;
1084	int error, retries = 0;
1085
1086retry:
1087	handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1088	if (IS_ERR(handle)) {
1089		error = PTR_ERR(handle);
1090	} else {
1091		int error2;
1092
1093		error = ext4_xattr_set_handle(handle, inode, name_index, name,
1094					      value, value_len, flags);
1095		error2 = ext4_journal_stop(handle);
1096		if (error == -ENOSPC &&
1097		    ext4_should_retry_alloc(inode->i_sb, &retries))
1098			goto retry;
1099		if (error == 0)
1100			error = error2;
1101	}
1102
1103	return error;
1104}
1105
1106/*
1107 * Shift the EA entries in the inode to create space for the increased
1108 * i_extra_isize.
1109 */
1110static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1111				     int value_offs_shift, void *to,
1112				     void *from, size_t n, int blocksize)
1113{
1114	struct ext4_xattr_entry *last = entry;
1115	int new_offs;
1116
1117	/* Adjust the value offsets of the entries */
1118	for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1119		if (!last->e_value_block && last->e_value_size) {
1120			new_offs = le16_to_cpu(last->e_value_offs) +
1121							value_offs_shift;
1122			BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1123				 > blocksize);
1124			last->e_value_offs = cpu_to_le16(new_offs);
1125		}
1126	}
1127	/* Shift the entries by n bytes */
1128	memmove(to, from, n);
1129}
1130
1131/*
1132 * Expand an inode by new_extra_isize bytes when EAs are present.
1133 * Returns 0 on success or negative error number on failure.
1134 */
1135int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1136			       struct ext4_inode *raw_inode, handle_t *handle)
1137{
1138	struct ext4_xattr_ibody_header *header;
1139	struct ext4_xattr_entry *entry, *last, *first;
1140	struct buffer_head *bh = NULL;
1141	struct ext4_xattr_ibody_find *is = NULL;
1142	struct ext4_xattr_block_find *bs = NULL;
1143	char *buffer = NULL, *b_entry_name = NULL;
1144	size_t min_offs, free;
1145	int total_ino, total_blk;
1146	void *base, *start, *end;
1147	int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1148	int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1149
1150	down_write(&EXT4_I(inode)->xattr_sem);
1151retry:
1152	if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1153		up_write(&EXT4_I(inode)->xattr_sem);
1154		return 0;
1155	}
1156
1157	header = IHDR(inode, raw_inode);
1158	entry = IFIRST(header);
1159
1160	/*
1161	 * Check if enough free space is available in the inode to shift the
1162	 * entries ahead by new_extra_isize.
1163	 */
1164
1165	base = start = entry;
1166	end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1167	min_offs = end - base;
1168	last = entry;
1169	total_ino = sizeof(struct ext4_xattr_ibody_header);
1170
1171	free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1172	if (free >= new_extra_isize) {
1173		entry = IFIRST(header);
1174		ext4_xattr_shift_entries(entry,	EXT4_I(inode)->i_extra_isize
1175				- new_extra_isize, (void *)raw_inode +
1176				EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1177				(void *)header, total_ino,
1178				inode->i_sb->s_blocksize);
1179		EXT4_I(inode)->i_extra_isize = new_extra_isize;
1180		error = 0;
1181		goto cleanup;
1182	}
1183
1184	/*
1185	 * Enough free space isn't available in the inode, check if
1186	 * EA block can hold new_extra_isize bytes.
1187	 */
1188	if (EXT4_I(inode)->i_file_acl) {
1189		bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1190		error = -EIO;
1191		if (!bh)
1192			goto cleanup;
1193		if (ext4_xattr_check_block(bh)) {
1194			EXT4_ERROR_INODE(inode, "bad block %llu",
1195					 EXT4_I(inode)->i_file_acl);
1196			error = -EIO;
1197			goto cleanup;
1198		}
1199		base = BHDR(bh);
1200		first = BFIRST(bh);
1201		end = bh->b_data + bh->b_size;
1202		min_offs = end - base;
1203		free = ext4_xattr_free_space(first, &min_offs, base,
1204					     &total_blk);
1205		if (free < new_extra_isize) {
1206			if (!tried_min_extra_isize && s_min_extra_isize) {
1207				tried_min_extra_isize++;
1208				new_extra_isize = s_min_extra_isize;
1209				brelse(bh);
1210				goto retry;
1211			}
1212			error = -1;
1213			goto cleanup;
1214		}
1215	} else {
1216		free = inode->i_sb->s_blocksize;
1217	}
1218
1219	while (new_extra_isize > 0) {
1220		size_t offs, size, entry_size;
1221		struct ext4_xattr_entry *small_entry = NULL;
1222		struct ext4_xattr_info i = {
1223			.value = NULL,
1224			.value_len = 0,
1225		};
1226		unsigned int total_size;  /* EA entry size + value size */
1227		unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1228		unsigned int min_total_size = ~0U;
1229
1230		is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1231		bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1232		if (!is || !bs) {
1233			error = -ENOMEM;
1234			goto cleanup;
1235		}
1236
1237		is->s.not_found = -ENODATA;
1238		bs->s.not_found = -ENODATA;
1239		is->iloc.bh = NULL;
1240		bs->bh = NULL;
1241
1242		last = IFIRST(header);
1243		/* Find the entry best suited to be pushed into EA block */
1244		entry = NULL;
1245		for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1246			total_size =
1247			EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1248					EXT4_XATTR_LEN(last->e_name_len);
1249			if (total_size <= free && total_size < min_total_size) {
1250				if (total_size < new_extra_isize) {
1251					small_entry = last;
1252				} else {
1253					entry = last;
1254					min_total_size = total_size;
1255				}
1256			}
1257		}
1258
1259		if (entry == NULL) {
1260			if (small_entry) {
1261				entry = small_entry;
1262			} else {
1263				if (!tried_min_extra_isize &&
1264				    s_min_extra_isize) {
1265					tried_min_extra_isize++;
1266					new_extra_isize = s_min_extra_isize;
1267					goto retry;
1268				}
1269				error = -1;
1270				goto cleanup;
1271			}
1272		}
1273		offs = le16_to_cpu(entry->e_value_offs);
1274		size = le32_to_cpu(entry->e_value_size);
1275		entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1276		i.name_index = entry->e_name_index,
1277		buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1278		b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1279		if (!buffer || !b_entry_name) {
1280			error = -ENOMEM;
1281			goto cleanup;
1282		}
1283		/* Save the entry name and the entry value */
1284		memcpy(buffer, (void *)IFIRST(header) + offs,
1285		       EXT4_XATTR_SIZE(size));
1286		memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1287		b_entry_name[entry->e_name_len] = '\0';
1288		i.name = b_entry_name;
1289
1290		error = ext4_get_inode_loc(inode, &is->iloc);
1291		if (error)
1292			goto cleanup;
1293
1294		error = ext4_xattr_ibody_find(inode, &i, is);
1295		if (error)
1296			goto cleanup;
1297
1298		/* Remove the chosen entry from the inode */
1299		error = ext4_xattr_ibody_set(handle, inode, &i, is);
1300		if (error)
1301			goto cleanup;
1302
1303		entry = IFIRST(header);
1304		if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1305			shift_bytes = new_extra_isize;
1306		else
1307			shift_bytes = entry_size + size;
1308		/* Adjust the offsets and shift the remaining entries ahead */
1309		ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1310			shift_bytes, (void *)raw_inode +
1311			EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1312			(void *)header, total_ino - entry_size,
1313			inode->i_sb->s_blocksize);
1314
1315		extra_isize += shift_bytes;
1316		new_extra_isize -= shift_bytes;
1317		EXT4_I(inode)->i_extra_isize = extra_isize;
1318
1319		i.name = b_entry_name;
1320		i.value = buffer;
1321		i.value_len = size;
1322		error = ext4_xattr_block_find(inode, &i, bs);
1323		if (error)
1324			goto cleanup;
1325
1326		/* Add entry which was removed from the inode into the block */
1327		error = ext4_xattr_block_set(handle, inode, &i, bs);
1328		if (error)
1329			goto cleanup;
1330		kfree(b_entry_name);
1331		kfree(buffer);
1332		b_entry_name = NULL;
1333		buffer = NULL;
1334		brelse(is->iloc.bh);
1335		kfree(is);
1336		kfree(bs);
1337	}
1338	brelse(bh);
1339	up_write(&EXT4_I(inode)->xattr_sem);
1340	return 0;
1341
1342cleanup:
1343	kfree(b_entry_name);
1344	kfree(buffer);
1345	if (is)
1346		brelse(is->iloc.bh);
1347	kfree(is);
1348	kfree(bs);
1349	brelse(bh);
1350	up_write(&EXT4_I(inode)->xattr_sem);
1351	return error;
1352}
1353
1354
1355
1356/*
1357 * ext4_xattr_delete_inode()
1358 *
1359 * Free extended attribute resources associated with this inode. This
1360 * is called immediately before an inode is freed. We have exclusive
1361 * access to the inode.
1362 */
1363void
1364ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1365{
1366	struct buffer_head *bh = NULL;
1367
1368	if (!EXT4_I(inode)->i_file_acl)
1369		goto cleanup;
1370	bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1371	if (!bh) {
1372		EXT4_ERROR_INODE(inode, "block %llu read error",
1373				 EXT4_I(inode)->i_file_acl);
1374		goto cleanup;
1375	}
1376	if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1377	    BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1378		EXT4_ERROR_INODE(inode, "bad block %llu",
1379				 EXT4_I(inode)->i_file_acl);
1380		goto cleanup;
1381	}
1382	ext4_xattr_release_block(handle, inode, bh);
1383	EXT4_I(inode)->i_file_acl = 0;
1384
1385cleanup:
1386	brelse(bh);
1387}
1388
1389/*
1390 * ext4_xattr_put_super()
1391 *
1392 * This is called when a file system is unmounted.
1393 */
1394void
1395ext4_xattr_put_super(struct super_block *sb)
1396{
1397	mb_cache_shrink(sb->s_bdev);
1398}
1399
1400/*
1401 * ext4_xattr_cache_insert()
1402 *
1403 * Create a new entry in the extended attribute cache, and insert
1404 * it unless such an entry is already in the cache.
1405 *
1406 * Returns 0, or a negative error number on failure.
1407 */
1408static void
1409ext4_xattr_cache_insert(struct buffer_head *bh)
1410{
1411	__u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1412	struct mb_cache_entry *ce;
1413	int error;
1414
1415	ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1416	if (!ce) {
1417		ea_bdebug(bh, "out of memory");
1418		return;
1419	}
1420	error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1421	if (error) {
1422		mb_cache_entry_free(ce);
1423		if (error == -EBUSY) {
1424			ea_bdebug(bh, "already in cache");
1425			error = 0;
1426		}
1427	} else {
1428		ea_bdebug(bh, "inserting [%x]", (int)hash);
1429		mb_cache_entry_release(ce);
1430	}
1431}
1432
1433/*
1434 * ext4_xattr_cmp()
1435 *
1436 * Compare two extended attribute blocks for equality.
1437 *
1438 * Returns 0 if the blocks are equal, 1 if they differ, and
1439 * a negative error number on errors.
1440 */
1441static int
1442ext4_xattr_cmp(struct ext4_xattr_header *header1,
1443	       struct ext4_xattr_header *header2)
1444{
1445	struct ext4_xattr_entry *entry1, *entry2;
1446
1447	entry1 = ENTRY(header1+1);
1448	entry2 = ENTRY(header2+1);
1449	while (!IS_LAST_ENTRY(entry1)) {
1450		if (IS_LAST_ENTRY(entry2))
1451			return 1;
1452		if (entry1->e_hash != entry2->e_hash ||
1453		    entry1->e_name_index != entry2->e_name_index ||
1454		    entry1->e_name_len != entry2->e_name_len ||
1455		    entry1->e_value_size != entry2->e_value_size ||
1456		    memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1457			return 1;
1458		if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1459			return -EIO;
1460		if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1461			   (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1462			   le32_to_cpu(entry1->e_value_size)))
1463			return 1;
1464
1465		entry1 = EXT4_XATTR_NEXT(entry1);
1466		entry2 = EXT4_XATTR_NEXT(entry2);
1467	}
1468	if (!IS_LAST_ENTRY(entry2))
1469		return 1;
1470	return 0;
1471}
1472
1473/*
1474 * ext4_xattr_cache_find()
1475 *
1476 * Find an identical extended attribute block.
1477 *
1478 * Returns a pointer to the block found, or NULL if such a block was
1479 * not found or an error occurred.
1480 */
1481static struct buffer_head *
1482ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1483		      struct mb_cache_entry **pce)
1484{
1485	__u32 hash = le32_to_cpu(header->h_hash);
1486	struct mb_cache_entry *ce;
1487
1488	if (!header->h_hash)
1489		return NULL;  /* never share */
1490	ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1491again:
1492	ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1493				       hash);
1494	while (ce) {
1495		struct buffer_head *bh;
1496
1497		if (IS_ERR(ce)) {
1498			if (PTR_ERR(ce) == -EAGAIN)
1499				goto again;
1500			break;
1501		}
1502		bh = sb_bread(inode->i_sb, ce->e_block);
1503		if (!bh) {
1504			EXT4_ERROR_INODE(inode, "block %lu read error",
1505					 (unsigned long) ce->e_block);
1506		} else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1507				EXT4_XATTR_REFCOUNT_MAX) {
1508			ea_idebug(inode, "block %lu refcount %d>=%d",
1509				  (unsigned long) ce->e_block,
1510				  le32_to_cpu(BHDR(bh)->h_refcount),
1511					  EXT4_XATTR_REFCOUNT_MAX);
1512		} else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1513			*pce = ce;
1514			return bh;
1515		}
1516		brelse(bh);
1517		ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1518	}
1519	return NULL;
1520}
1521
1522#define NAME_HASH_SHIFT 5
1523#define VALUE_HASH_SHIFT 16
1524
1525/*
1526 * ext4_xattr_hash_entry()
1527 *
1528 * Compute the hash of an extended attribute.
1529 */
1530static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1531					 struct ext4_xattr_entry *entry)
1532{
1533	__u32 hash = 0;
1534	char *name = entry->e_name;
1535	int n;
1536
1537	for (n = 0; n < entry->e_name_len; n++) {
1538		hash = (hash << NAME_HASH_SHIFT) ^
1539		       (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1540		       *name++;
1541	}
1542
1543	if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1544		__le32 *value = (__le32 *)((char *)header +
1545			le16_to_cpu(entry->e_value_offs));
1546		for (n = (le32_to_cpu(entry->e_value_size) +
1547		     EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1548			hash = (hash << VALUE_HASH_SHIFT) ^
1549			       (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1550			       le32_to_cpu(*value++);
1551		}
1552	}
1553	entry->e_hash = cpu_to_le32(hash);
1554}
1555
1556#undef NAME_HASH_SHIFT
1557#undef VALUE_HASH_SHIFT
1558
1559#define BLOCK_HASH_SHIFT 16
1560
1561/*
1562 * ext4_xattr_rehash()
1563 *
1564 * Re-compute the extended attribute hash value after an entry has changed.
1565 */
1566static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1567			      struct ext4_xattr_entry *entry)
1568{
1569	struct ext4_xattr_entry *here;
1570	__u32 hash = 0;
1571
1572	ext4_xattr_hash_entry(header, entry);
1573	here = ENTRY(header+1);
1574	while (!IS_LAST_ENTRY(here)) {
1575		if (!here->e_hash) {
1576			/* Block is not shared if an entry's hash value == 0 */
1577			hash = 0;
1578			break;
1579		}
1580		hash = (hash << BLOCK_HASH_SHIFT) ^
1581		       (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1582		       le32_to_cpu(here->e_hash);
1583		here = EXT4_XATTR_NEXT(here);
1584	}
1585	header->h_hash = cpu_to_le32(hash);
1586}
1587
1588#undef BLOCK_HASH_SHIFT
1589
1590int __init
1591init_ext4_xattr(void)
1592{
1593	ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1594	if (!ext4_xattr_cache)
1595		return -ENOMEM;
1596	return 0;
1597}
1598
1599void
1600exit_ext4_xattr(void)
1601{
1602	if (ext4_xattr_cache)
1603		mb_cache_destroy(ext4_xattr_cache);
1604	ext4_xattr_cache = NULL;
1605}
1606